Copyright @ IFAC Control Applications of Optimization, SI. Petersburg, Russia, 2000 A NEW STRATEGY FOR THE OPTIMIZATION LEVEL IN MODEL PREDICTIVE CONTROL Maria Jose Arbiza, Jose Alberto Bandoni, Jose Luis Figueroa l Planta Piloto de Ingenieria Quimica (PLAPIQUJ - UNS - CONICET) Camino La Carrindanga Km. 7. - 8000 - Bahia Blanca- ARGENTINA Tel. (054) 2914861700 Fax. (054) 291 4861600 e-mail: cofiguer@criba.edu.ar Abstract: The desired operating point in Model Predictive Control is detennined by a local steady-state optimization, which may be based on an economic objective and a linear model. In this paper, we incorporate the solution of a back-off problem to obtain a hierarchical scheme that ensures feasible operation in despite of possible disturbances. In particular, we use a relaxed version of the classical back-off algorithm, assuming linear models for the process and the constraints and a quadratic objective function. Copyright ©2000 IF A C Keywords: model based control, dynamic programming, mathematical programming, multilevel control, constraints. 1. INTRODUCTION In the last years the Model Predictive Control (M PC) has been considered by researchers and practitioners as one of the most important developments in control. The credit for its remarkable success is generally given to several industrialists, who outlined the basic algorithms and argued about its potential for industrial applications. The well publicized success of MPC in the process industries has fueled and in many ways impulsed the research in academy (Lee and Cooley, 1997; Qin and Badgwell, 1997). In general, MPC refers to a class of computer implemented mathematical algorithms that control the future behavior of a plant through the use of an I Also in Dpto. de Ing. Electrica - UNS - Avda. Alem 1253 - (8000) Bahia Blanca - ARGENTINA 17 explicit process model. At each control interval the MPC algorithm computes in an open-loop mode a sequence of adjustments on manipulated variables, in order to optimize the future plant behavior under process constraints. The first input in the optimal sequence is injected into the plant, and the entire optimization is repeated at subsequent control intervals. In the modem processing plants the MPC controller is part of a multi-level hierarchy of control functions (Qin and Badgwell, 1997), as it is illustrated in Figure 1. Similar hierarchical structures have been described by several other authors (Richalet et ai, 1978; Prett and Garcia, 1988). The second stage of this hierarchy (the unit optimizer) computes an optimal steady-state point and passes this to the dynamic constraint control system for its implementation. This desired operating point is usually detennined by a local